NVIDIA Pascal Thread

Page 82 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Kris194

Member
Mar 16, 2016
112
0
0
GTX 1070 for 500$? It's either very powerful or it's serious milking. If it doesn't beat Titan X by 30% then there's no way I will buy this shit.
 
Last edited:

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
No one is going to buy a 1070 for $500. Their yields must be horrible if that's the number they came up with.
 
Feb 19, 2009
10,457
10
76
1070 was leaked to have similar to 980Ti 3dmark graphics score.

$499 or around there for a 980Ti class GPU with much less power usage and 8GB vram won't sell?

I think you guys will be very wrong. All the 970 3.5GB vram gimped gamers will love that 8GB and extra performance.

Let's apply a few things we know about Pascal, it's got a better suited uarch for compute heavy games, it's got a GCN-like layout and warp optimization, so next-gen console ports will not be losing performance.

Now if its got similar to the 980Ti in 3dMark, it's actual gaming performance in modern games, will be higher than the 980Ti by a fair margin.

Edit: Basically we may see the 1070 as being ~980Ti or a little below it in older games, but the 1070 could end up ~20% faster in modern titles. This means the 1080 will be ~35% faster in modern titles. That's enough to justify it's price, easily.
 
Last edited:

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
1070 was leaked to have similar to 980Ti 3dmark graphics score.

$499 or around there for a 980Ti class GPU with much less power usage and 8GB vram won't sell?

I think you guys will be very wrong. All the 970 3.5GB vram gimped gamers will love that 8GB and extra performance.

Let's apply a few things we know about Pascal, it's got a better suited uarch for compute heavy games, it's got a GCN-like layout and warp optimization, so next-gen console ports will not be losing performance.

Now if its got similar to the 980Ti in 3dMark, it's actual gaming performance in modern games, will be higher than the 980Ti by a fair margin.

I don't think the average enthusiast gamer can afford $500 for a gpu. They sold so many 970s because they were $330 and performed almost like a $650 780 Ti.
 

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
If $500 midrange chips becomes the norm I'm done with PC gaming once my 970 starts sucking and will buy a Playstation next generation.
 
Feb 19, 2009
10,457
10
76
I don't think the average enthusiast gamer can afford $500 for a gpu. They sold so many 970s because they were $330 and performed almost like a $650 780 Ti.

How about this scenario?

GP106 will be ~$349 performing around the 980 mark. With cut down variants being much cheaper, around the 970 performance level.

The cut-down variants may not even need a 6x pin, ~75W. Would that sell? Heck yes.
 

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
Chiphell: Geforce GTX 1080 / GTX 1070 possible price revealed



In comparison, here's Geforce GTX 980 Ti and Geforce GTX 980 in Newegg.tw:
www.newegg.com.tw/item?itemid=533505 - NT $ 27000
www.newegg.com.tw/item?itemid=129836 - NT $ 20000

And US Newegg.com:
www.newegg.com/Product/Product.aspx...980Ti_6GB_ZT-90505-10P-_-14-500-379-_-Product
www.newegg.com/Product/Product.aspx?Item=N82E16814127840

An educated guess:
- Geforce GTX 1080: $649
- Geforce GTX 1070: $499

Is this guy at chiphell any kind of reasonable source?
 

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
How about this scenario?

GP106 will be ~$349 performing around the 980 mark. With cut down variants being much cheaper, around the 970 performance level.

The cut-down variants may not even need a 6x pin, ~75W. Would that sell? Heck yes.

The lower midrange chip for $350 really sounds perverse. I wouldn't figure Nvidia would be able to get away with that until after they had killed off AMD.
 

iiiankiii

Senior member
Apr 4, 2008
759
47
91
The lower midrange chip for $350 really sounds perverse. I wouldn't figure Nvidia would be able to get away with that until after they had killed off AMD.

They can get away with it if AMD is also milking 14nm because performance is starting to hit a brick wall. No more easy node shrinks from here on out.
 
Feb 19, 2009
10,457
10
76
The lower midrange chip for $350 really sounds perverse. I wouldn't figure Nvidia would be able to get away with that until after they had killed off AMD.

At one point they got to 82 - 18 marketshare. So in effect, it's a monopoly.

NV figures they have the mindshare, lots of gamers won't touch AMD regardless so they can price it at whatever the want. Right?

We all know prices will keep going up, as low level dGPU get killed off by the encroaching APU/iGPU. What was $200 (960) will be $300. etc
 

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
At one point they got to 82 - 18 marketshare. So in effect, it's a monopoly.

NV figures they have the mindshare, lots of gamers won't touch AMD regardless so they can price it at whatever the want. Right?

We all know prices will keep going up, as low level dGPU get killed off by the encroaching APU/iGPU. What was $200 (960) will be $300. etc

I hope not. Is there any reason to trust that chiphell source?
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Yep, the funny part is many said this at the time, but the ADF denounced this simple fact. However, they now say the 680 was midrange. It's really hard to get the facts right around here because so many manipulate the truth for silly purposes.

Firstly, there was major negative feedback regarding early launch prices for the entire HD7000 stack on this very forum, even before GTX670/680 showed up. Secondly, after 670/680 showed up, again on this very forum, there was a massive backlash and many months of discussions about how NV doubled the prices of cards by pricing mid-range products at flagship pricing levels. Many guys who bought NV criticized 680 for delivering the worst "flagship" generation increase ever (over the 580), hence its $499 pricing wasn't justified even in their own eyes. But, most of the market complained and still purchased $350-550 7950/7970/670/680 cards. That sets off a new precedent for AMD/NV. Thirdly, this new strategy was repeated with Maxwell with NV pushing the price of 560Ti's successor from $499 of 680 to $549 for the 980. The term 'bi-furcating' a generation became a common way to describe this new way to release next gen products. The majority of 'ADF' as you put it called it as it was. The difference was that many of those same gamers also mined with AMD cards, which meant the $550 price on 7970 and $550 on 290X weren't a deterrent as those cards were essentially paid off via crypto-currency. That doesn't mean many on here didn't criticize the direction the entire GPU industry was heading, as early as 1-2Q 2012.

Whatever, I am not sweating it. As long as crypto-currency continues into its 8th year, I'll be well on my way to free Vegas. If NV wants to milk its customer base with $500 1070 and $650 1080 (mid-range parts), that's their right. At the end of the day, NV and AMD are going to price what the market can bear and if people want to pay $500-650 for mid-range products, well...congrats AMD/NV! Sooner or later crypto-currency won't sustain free GPU upgrades and then it won't be fun anymore paying $550 for mid-range next gen parts.

How about 980Ti + 20% at 150W?

$649 would slot right in. Incremental improvements with a major power efficiency gains. Hello Intel's playbook!

The full GP104 with GDDR5X, 1080Ti, $899?? ~Titan X + 30%?

After-market $420 670 beat 580 by 30% at 1600p.



So when 1080 beats 980Ti by the same 20-30%, it should cost $600-650 because.....inflation and Huang's next gen Ferrari/Tesla stable replacements.

God, $650 for the midrange chip? If that happens I'm definitely waiting for Vega.

Ok, and if 1080 is $550-650 and Vega 10/11 is even faster, what makes you think AMD won't price it for at least $650? AMD was perfectly content pricing the Nano for $649 and Fury X for $649 despite after-market 980Ti cards demolishing them by 25-30% once fully overclocked! Ouch. The old management of AMD would have never priced a card which is definitively slower and has less VRAM to go head on against the 980Ti. That's just market share suicide. With new AMD, it's now way too hard to predict but I think Fiji pricing and AMD raising market prices on 290/290X with 390/390X replacements is a good indication that AMD is no longer interested in $369-379 5870/6970 style pricing strategies for their larger die product lines.

Is this guy at chiphell any kind of reasonable source?

Could also be a strategic leak by NV to scare the market with $499 1070 and $649 1080 prices to release them later at $429 and $599 and suddenly they are much cheaper than expected.

At one point they got to 82 - 18 marketshare. So in effect, it's a monopoly.

NV figures they have the mindshare, lots of gamers won't touch AMD regardless so they can price it at whatever the want. Right?

You hit the nail on the head there. With declining volume unit sales of CPUs and discrete GPUs, the best way to continue with revenue and profit growth is to increase gross margins by lowering costs (smaller die sizes), while raising prices (higher ASPs). Bifurcating a generation, doing fancy renaming by calling x60 series cards x70/80, VRAM gimping, releasing various cut-down flagships until finally releasing a fully unlocked flagship -- all of this isn't a coincidence. We should be lucky we enjoyed many years of $200-250 next gen cards.

Starting in 2012, mid-range became high-end and high-end became Enthusiast.



As long as NV's loyal user-base keeps buying, there is no reason to stop raising prices esp. since AMD is MIA with Vega 10/11 and Fury X cannot even beat an after-market 980Ti, so....
 
Last edited:
Feb 19, 2009
10,457
10
76
I hope not. Is there any reason to trust that chiphell source?

No, there's little reason to trust most leaks.

It falls down to whether you think it's a logical scenario.

Do you remember when the 970 was launched at $330 and we all thought it was cheap? We expected to be around $399.

Think back why.. AMD at the time at 37% marketshare on a rising trajectory. The 970 as I said back then, was the killing blow and it did the most damage.

Now that NV is in a dominating position...

If you were in JHH's shoes, what would you do to maximize revenue & profits?

ie, price GP104 low to gain another 5% marketshare will result in less revenue/profit. Rather, jacking up the prices/margins will potentially see a small decline in marketshare, but a higher revenue/profit.
 

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
No, there's little reason to trust most leaks.

It falls down to whether you think it's a logical scenario.

Do you remember when the 970 was launched at $330 and we all thought it was cheap? We expected to be around $399.

Think back why.. AMD at the time at 37% marketshare on a rising trajectory. The 970 as I said back then, was the killing blow and it did the most damage.

Now that NV is in a dominating position...

If you were in JHH's shoes, what would you do to maximize revenue & profits?

ie, price GP104 low to gain another 5% marketshare will result in less revenue/profit. Rather, jacking up the prices/margins will potentially see a small decline in marketshare, but a higher revenue/profit.

If I was in his shoes I would try to kill off AMD for good this generation with another 970 level deal and then jack the prices way up. I don't think AMD could survive another 970 in the gpu market.
 
Feb 19, 2009
10,457
10
76
You hit the nail on the head there. With declining volume unit sales of CPUs and discrete GPUs, the best way to continue with revenue and profit growth is to increase gross margins by lowering costs (smaller die sizes), while raising prices (higher ASPs). Bifurcating a generation, doing fancy renaming by calling x60 series cards x70/80, VRAM gimping, releasing various cut-down flagships until finally releasing a fully unlocked flagship -- all of this isn't a coincidence. We should be lucky we enjoyed many years of $200-250 next gen cards.

Starting in 2012, mid-range became high-end and high-end became Enthusiast.


Well said, with the included references to boot. I was too lazy to get the links to explain why it's expected for a rising in prices.

NV doesn't have much growth opportunities in dGPU because they already captured most of the market. So the only way to grow revenue and importantly, profits, is to maximize margins.

This means lower cost (smaller die), at a higher price.

It is inconceivable for GP100 being a 1080Ti because NV can milk each for more than 10x the price as a Tesla.

What are their options? Unless you believe in the fairytale GP102... NV only has GP104.

How to max their profit from GP104? Tier up each SKU. Instead of 680, 670, and 660Ti, it's 680Ti, 680, 670 etc.

If the performance gap is not enough of an incentive for upgrades (I think it will be enough in newer games, DX12 etc), planned obsolescence via drivers (Kepler anyone?) & GameWorks will do the job.
 
Feb 19, 2009
10,457
10
76
If I was in his shoes I would try to kill off AMD for good this generation with another 970 level deal and then jack the prices way up. I don't think AMD could survive another 970 in the gpu market.

AMD cannot be killed off by a dGPU attack. It's a small portion of their revenue.

Their entire computing and graphics division earn ~50% revenue, making a loss this quarter.

Semi-custom and others (profitable) made up the other 50%.

Within that 50%, HPC/Firepros/Apple from AMD is actually doing quite well. AMD's CPU tanks and GPU is on life-support.

If NV price aggressive Pascal to gain another 10% marketshare to effectively kill off AMD dGPUs, it gains NV very little and AMD will STILL be around.

They just got some major wins, 3x next-gen consoles on top of current consoles. Apple refreshes Polaris 11/10. IP wins (free $$) with royalties for China-based x86 designs. AMD will be just fine even if they sold no dGPUs. Really. The IP win alone is worth more revenue than their dGPU sales per quarter.

So if you were in JHH shoes and tried to actively price your Pascal to kill AMD dGPU, you would lose, your revenue/profit won't be as high and your shares will drop. Shareholders will not be happy with you.
 

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
AMD cannot be killed off by a dGPU attack. It's a small portion of their revenue.

Their entire computing and graphics division earn ~50% revenue, making a loss this quarter.

Semi-custom and others (profitable) made up the other 50%.

Within that 50%, HPC/Firepros/Apple from AMD is actually doing quite well. AMD's CPU tanks and GPU is on life-support.

If NV price aggressive Pascal to gain another 10% marketshare to effectively kill off AMD dGPUs, it gains NV very little and AMD will STILL be around.

They just got some major wins, 3x next-gen consoles on top of current consoles. Apple refreshes Polaris 11/10. IP wins (free $$) with royalties for China-based x86 designs. AMD will be just fine even if they sold no dGPUs. Really. The IP win alone is worth more revenue than their dGPU sales per quarter.

So if you were in JHH shoes and tried to actively price your Pascal to kill AMD dGPU, you would lose, your revenue/profit won't be as high and your shares will drop. Shareholders will not be happy with you.

Damn man, you're taking away all the excitement I had for this new generation haha.
 
Feb 19, 2009
10,457
10
76
Damn man, you're taking away all the excitement I had for this new generation haha.

GP104 and Polaris aren't really the exciting stuff anyway. I was looking forward to something more powerful, Vega.

If they want us to pay a premium, may as well get awesome performance. I would prefer to pay $1K for a huge chip with HBM2 than $649 for a mid-range with GDDR5.
 

SteveGrabowski

Diamond Member
Oct 20, 2014
7,428
6,157
136
GP104 and Polaris aren't really the exciting stuff anyway. I was looking forward to something more powerful, Vega.

If they want us to pay a premium, may as well get awesome performance. I would prefer to pay $1K for a huge chip with HBM2 than $649 for a mid-range with GDDR5.

$600 is about the most I could ever see myself paying for a gpu, and it would have to be something extremely impressive. I hope you're wrong and that the market won't sustain those kind of prices. Even with a $340 970 (what I paid in 2014) I already feel like my computer is a money pit.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
How to max their profit from GP104? Tier up each SKU. Instead of 680, 670, and 660Ti, it's 680Ti, 680, 670 etc.

If the performance gap is not enough of an incentive for upgrades (I think it will be enough in newer games, DX12 etc), planned obsolescence via drivers (Kepler anyone?) & GameWorks will do the job.

That was my fear earlier. $299 660Ti becomes 1070 and price goes up $100-150. Recall 660Ti actually tied or even beat GTX580 (that's a Titan X predecessor!). So now if GTX1060Ti aka 1070 matches or slightly beats 980Ti, it actually looks like a great value at $399-449 doesn't it? The problem is 660Ti accomplished this task at $299. Release 1060 or 1060Ti at $299 and it destroys the GTX970. Suddenly that also looks like bringing a ton of value on the table and one of the best 1060 cards in years (but that's because 960 was not a true x60 series card). This marketing strategy is utterly brilliant but in reality it's raising prices across tiers again, just changing the names and introducing a 3rd lower tier GP104.

$600 is about the most I could ever see myself paying for a gpu, and it would have to be something extremely impressive. I hope you're wrong and that the market won't sustain those kind of prices. Even with a $340 970 (what I paid in 2014) I already feel like my computer is a money pit.

You know 970 OC is still a very good performer for 1080p. You don't need to upgrade to a $500-650 GP104 card if your current card performs to your satisfaction. There are a lot of games coming out where Ultra and Medium/High look 99% identical and performance delta is like 50-100% in FPS.

Sometimes, it's so bad that lowest and highest require a magnifying glass.

Low vs. High
Low vs. High 2
Low vs. High 3
 
Last edited:

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
None of this makes any sense... The 670 is a cut down GK104, the full dies were going to be sold either way. The base clock for the 670 is 915MHz. I'm not sure where you got 1.1-1.2GHz, as even the 680's base clock is 1006MHz.

Base clocks mean nothing. [H]'s first 680 boosted to over 1300MHz.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Base clocks mean nothing. [H]'s first 680 boosted to over 1300MHz.

Titan X's official boost is only 1075mhz but in the real world it achieved 1215mhz. That means on paper it looks nice to pretend a 1500mhz Titan X has 39.5% overclocking headroom.

That is quite a large chip. Anyone here to approx die size?

Looks smaller than GM204. I'd guess 300-325mm2 max.

 
Last edited:

JDG1980

Golden Member
Jul 18, 2013
1,663
570
136
That is quite a large chip. Anyone here to approx die size?

It's pretty small for a '4'-series Nvidia chip. Probably at or slightly below 300m^2. That means the transistor count is going to be about the same as GM200 (assuming 2:1 scaling for the node shrink), so any gains will have to come from clock speeds and/or architecture. I still don't see how Nvidia is going to pull off enough performance gains over GM200 to justify the prices we've seen leaked.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |